JP3161942B2 - Translation machine translation device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a translation machine translation apparatus for arranging and outputting translated words near each language in an original sentence.

[0002]

2. Description of the Related Art Conventionally, when translating a sentence written in a certain language into another language, it is sometimes difficult to establish a correspondence between the translated sentence and the input sentence in a long sentence in which the input sentence extends over a plurality of lines. In particular, when the sentence structures are significantly different, such as in English and Japanese, the phenomenon that "the translation of the phrase on the first line of the input sentence exists on the third line in the translated sentence" is often seen.

The current machine translation system presumes post-editing of a translated sentence, and the inability to find the correspondence between words at a glance in the process of comparing an input sentence and a translation result at a glance is due to the translation accuracy. Burden on the user.

When reading or reading a foreign language sentence quickly, or when a person who can understand English to a certain extent or an input sentence has a somewhat simple structure, a one-sentence unit having the above-described problem is generated. If a translation is assigned only to a difficult English word that you do not know, it may be more convenient to read and understand the sentence yourself.

[0005] As a conventional technology to meet such demands,
JP-A-6-243162 discloses a "translation support device" and JP-A-6-325081 discloses a "machine translation device".

First, Japanese Patent Laid-Open No. Hei 6-243162 relates to a translation support apparatus for arranging and outputting a translated word in the vicinity of each language in an original sentence and clarifying the relationship between the original word and the translated word. As shown in FIG. 12, the translation support apparatus stores dictionary data for converting Japanese sentence data input from the key input unit 21 and stored in the sentence memory 22 into kanji, and also stores English document data. A dictionary memory 23 for storing translated word data for translating a word, a translation-free word memory 24 and a document memory 22 for storing original words (English words) that need not be translated among the original words read from the document memory 22 during the translation support processing. And a register 25 formed from a translated word register 25b for storing the translated word read from the.

The operation of the translation support device 26 will be described below.

For example, "I will purchase the restaura
It is assumed that the English sentence “nt from my uncle.” is input from the key input unit 21 and stored in the document memory 22. When a translation is instructed by this key operation, the CPU 27
The character data in the document memory 22 is stored in the source language register 25a, and the number of digits of the source language is counted to obtain the source word digit register 25c.
To be stored in the original language register 25a.
Is searched for whether or not the original word data stored in the translation-necessary word memory 24 is stored, and it is determined whether or not it is necessary to add a translation word to the original word. For example, "I", "will", "
Assuming that the "," from "," my "are stored in the translation unnecessary word memory 24, the remaining" purchase "," restaurant "," unc "
Since it is necessary to add a translated word to "le", the corresponding translated word is searched from the dictionary memory 23, and when there is a translated word,
The translated word is read from the dictionary memory 23 and the translated word register 25
b, the number of digits of the translated word is calculated and stored in the translated word digit register 25d. Thereafter, display processing is performed according to the number of digits of the original language and the translated word, and the display is performed on the display unit 29 via the display memory 28. At this time, as shown below, translated words are assigned in units of words for easy viewing.

[0009] Next, Japanese Unexamined Patent Publication No. 6-325081 entitled "Machine Translator"
Will be described.

[0010] This Japanese Patent Application Laid-Open No. Hei 6-325081 selects and presents only information of a target original language such as a part of speech or a translated word.
A machine translation device that facilitates reading of target language. As shown in FIG. 13, the machine translation device includes an input unit 31 for inputting a first original sentence, a morphological analysis unit 33 for performing a morphological analysis using an analysis dictionary 32, and a A part-of-speech estimating unit 35 for finding an optimal part-of-speech corresponding to each word of the first original sentence using the chain establishment table 34, and an output form determining unit for creating an output in which information of the first language and the information of the second language are associated with each other. And output means 37 for outputting to an output device.

The operation of the machine translation device 38 will be described below.

For example, "I will book the restaurant fo
If the English sentence "r my uncle." is input by the input means 31, it is divided into morphemes by morphological analysis, and the dictionary is looked up for each morpheme. As a result, in a multi-speech word such as English, as shown below, A plurality of candidates are set for each morpheme.

Word Part-of-speech candidate I Pronoun will auxiliary verb / noun book noun / verb the article restaurant noun for preposition / connective my pronoun / exclamation uncle noun In the above example sentence, "will", "book", "for", "my" Has multiple parts of speech and is ambiguous. Therefore, the part of speech estimation means 3
In step 5, the part-of-speech connection probability is estimated, the appearance probability of the part-of-speech sequence is calculated based on the connection probability, and the part-of-speech sequence having the highest probability is selected. In this conventional example, a part-of-speech estimating means 35 uses a trigram model that limits the connection probability to the two preceding parts-of-speech, and further takes into account the relative part-of-speech appearance probability in words.
Now, the part of speech of the input sentence is estimated by the part of speech estimation means 35 as shown below. In particular, the part of speech of "book" can be estimated as a "verb" because the part of speech immediately before is an "auxiliary verb".

The word part of speech candidate I pronoun will [auxiliary verb] / noun book noun / [verb] the article restaurant noun for [preposition] / conjunction my [pronoun] / exclamation uncle noun [] The information of the second language corresponding to the part of speech may be obtained from the dictionary, and the output form determining means 36 may add the information of the second language in correspondence with the position of the word of the first language. At this time, in order to suppress frequent output of well-known contents of the contents of the second language information, an output suppression column is provided and marked in the dictionary. Also, for a heading having a plurality of usages, if the priority is numerically entered for each usage, it is possible to suppress output of only some usages. For example, "book" in the example sentence is suppressed when it is "noun (book)", and is output when it is "verb (reserve)". With the above processing, the output result for the example sentence is as follows.

[0015]

[0016]

However, the above-mentioned conventional translation support device 26 and machine translation device 38 have the following problems (a) and (b).

(A) The accuracy of the translated word is poor.

The translation support device 26 of Japanese Patent Laid-Open No. Hei 6-243162 assigns the same translation (the translation stored first or the translation with the highest priority) every time, regardless of the words before and after the sentence. For example, the most common interpretation of the word “book” is that the part of speech is “noun” and the translation is “book”, so even the following example sentence would give an incorrect translation of “book”.

[0019] On the other hand, in the machine translator 38 disclosed in Japanese Patent Laid-Open No. Hei 6-325081, as described above, the part-of-speech with the immediately preceding two words is checked. "Can be given the correct translation.

However, in the machine translation apparatus 38 disclosed in Japanese Patent Laid-Open No. 6-325081, even if, for example, parsing is used, only the part of speech can be specified. You cannot choose. For example, "take" is an English word with various meanings (translations) as follows, but take (verb) take, take, take, take, ride, eat, subscribe, need, receive, Take ... It is not possible to estimate which part of speech is to be selected simply by estimating the part of speech as a "verb". In the input sentence shown below, the most common translation of "take" is used, but in practice, if the object means "person", the translation of "take" is appropriate .

[0021] As described above, a user who reads these two input sentences to which an incorrect translation has been assigned will misunderstand. Of course, JP-A-6-243162 and JP-A-6-325
Although both translations can output a plurality of candidate translations, it imposes a burden on the user to select a translation.

(B) No translation is given only to words that are really needed.

It is sufficient for a user to assign a translation only to a word that he does not know. vice versa,
Even if a translation is given to a word that "knows", it is often disturbing. JP-A-6-243162 and JP-A-6-3
In Japanese Patent Publication No. 25081, “words not requiring translation”, “setting of fields not requiring output suppression” , and “priority to translations” are provided in advance to limit the words to which translations are assigned. While these "translation-unnecessary words" are fixed in the system, "knowing" or "don't know" a certain word greatly differs depending on the user. In other words, even words that are not known to English “beginners” (words for which translations are desired) are sufficiently considered to be words that are known to “experts”. As in JP-A-6-325081, in a system in which the translation is determined to suppress the use of any user, a really useful translation can not be assigned.

As described above, it is confusing that (a) the accuracy of the translated word is poor or (b) there are many words unnecessary for translation, and it is possible to efficiently and easily read a foreign language sentence quickly and roughly. There was a problem that can not be.

An object of the present invention is to solve the above-mentioned conventional problems, and an object of the present invention is to provide a translation machine translation apparatus capable of efficiently and easily reading an input sentence in a foreign language efficiently and easily. .

[0026]

SUMMARY OF THE INVENTION A translation machine translation apparatus according to the present invention translates grammatical information such as part of speech of a word and a translated word.
At the same time, the translation level is assigned to each translation.
And the input sentence is divided into word strings,
Grammar information such as the part of speech for each word in the word string
And get the translated words and solve the tense, person, and number of each word.
A dictionary lookup / morphological analysis means for analyzing, and the dictionary lookup / form
According to the grammatical information obtained by the morphological analysis means, the word sequence
Syntax to determine input sentence structure such as dependency relation between words
Analysis means, and the input sentence structure determined by the syntax analysis means
Into a sentence structure for the translated sentence,
Syntax conversion means for checking and selecting the translation for the word
And the order of the translated words obtained by the syntax conversion means is determined.
Translation generator that generates translations by adjusting
A translation machine translation device having a step and
Level information such as difficulty, frequency or importance of each word
Translation that can specify the translation level set based on the
Specified by the translation level setting means and the translation level setting means
Translation level and translation level in the word dictionary
Translation to determine whether to add a translation word based on the relationship with
Output determination means and translation processing up to the translation sentence generation means.
After performing, the translated word determined by the translation output determination means
On the other hand, each word of the input sentence is
A source that prepares the translated word together with the input sentence in an easy-to-view format and outputs it.
And a translation output unit.
The translation means specifies the translation specified by the translation level setting means.
Between the translation level and the translation level in the word dictionary
And whether or not to add a translation by
The syntax structure of the sentence obtained during the parsing process,
Character type configuration of the translated word obtained, obtained during the conversion process
Whether the translated word is specified in the field, and is the same
At least one of the distances between
Use information as linguistic information and assign translations based on the linguistic information
It is characterized in that it is determined whether or not to do so, whereby the above object is achieved.

Further, preferably, the translation machine translation apparatus of the present invention is arranged such that the distance between the positions at which the same translation is provided is immediately preceding.
It is characterized by the distance between the same translated words given , thereby achieving the above object.

Further, the translation machine translation apparatus of the present invention
When storing grammatical information such as part of speech for words and translations
Both are word dictionaries to which a translation level is assigned for each of the translated words
And the input sentence is divided into word strings, and the word
Get grammatical information such as part of speech and translations for each word in the column
And a dictionary that analyzes the tense, person, and number of each word
Lookup / morpheme analysis means and the dictionary lookup / morpheme analyzer
The relationship between the words in the word string according to the grammatical information obtained in the column
Parsing means to determine input sentence structure such as receiving relation
And the translated sentence structure determined by the parsing means.
Into a sentence structure for
Syntactic conversion means for selecting the translation for the word;
Determine the order of the translated words obtained by the syntax conversion means, and
Translation generation means for generating a translation by adjusting
A translation machine translation device, wherein a user
Based on level information such as difficulty, frequency or importance
Translation level setting that can specify the translation level set
Translation means specified by the translation level setting means.
Between the translation level and the translation level in the word dictionary
Translation output determination method that determines whether to add a translated word based on
After performing the translation processing up to the step and the translation sentence generation means
Next, the translated word determined by the translation output
Correspondence between each word in the literary sentence and its translation is entered into the translation.
Supports original language translations that are output in an easy-to-read format along with the power sentences
Output means, and the translation output determination means,
The translation level specified by the translation level setting means;
Translated words according to the translation level in the word dictionary
To determine whether or not to add
Sentence structure obtained during processing, obtained during the conversion process
The character type configuration of the translated word, the translated word obtained during the conversion process is
Whether or not the field has been designated, and
Information on at least one of the distances between
Whether to add a translation based on the linguistic information
Is determined, and the same is obtained during the syntax conversion process.
The number of times a translated word is assigned exceeds a predetermined standard within a predetermined range of the document.
If the name is changed, it is noted that the same translation will not be added after that.
The purpose is achieved.
It is .

[0029] In addition, preferably, translation pretend machine translation equipment of the present invention, the swing level translation to be applied to in the word dictionary
Is the first of the importance marks attached to the normal dictionary.
Information and the number of word characters constituting each word of the input sentence.
Second information, third information on the type of part of speech,
For dictionary types such as dictionaries, user dictionaries, or specialized dictionaries
Composed of at least one of the fourth information
Setting a translation level based on the first information
In some cases, out of multiple dictionaries, how many dictionaries have importance marks
Has been given, and how important the importance mark is
Set based on the necessity mark and based on the second information
When setting the translation level, the number of words
Whether it exceeds the number of characters, or the prefix of each word and
The meaning of each word is deduced from the configuration of each word such as a suffix.
Set based on whether it is measurable or not, and based on the third information
When setting the translation level, level
Low, and proper nouns and auxiliary verbs can be translated
And set the translation level based on the third information.
The word registered in the basic dictionary
Lower, and it is registered in the user dictionary or specialized dictionary.
Words in the word dictionary by raising the word level
It is characterized by setting the translation level to be given to
Therefore, the above object is achieved .

[0030]

[0031]

[0032]

[0033]

[0034]

[0035]

According to the present invention, an input sentence is divided into morpheme strings of words, and a word dictionary is searched for each word.
Analyze the tense, person, and number of the word. The input sentence structure such as the dependency between words in the morpheme sequence is determined, converted into the sentence structure for the translated sentence, and checked for semantic consistency to select a translated word for the word. Further, the arrangement of the translated words is determined, the translation is generated by adjusting the utilization change, and the translated sentence and the input sentence are arranged in an easy-to-view format.
Thereafter, the correspondence between each word of the input sentence and the translated word is output. In this way, after performing the morphological analysis, syntactic analysis, syntactic conversion, and translation processing for generating the translated sentence, translation is performed by associating each word of the input sentence with the translated word, so that a more accurate translated word is obtained. Can be output, the user can correctly understand the input sentence, and can quickly and easily read the input sentence in a foreign language efficiently and easily.

The translation level set by the translation level setting means set based on level information such as the degree of difficulty, frequency and importance of each word, and the level of difficulty, frequency and importance of each word Whether or not to add a translation is determined based on the relationship with the translation level in the word dictionary set based on level information such as the degree. Is not added, and it is possible to add more accurate and appropriate translations to the minimum required words desired by the user, and the user can efficiently read or roughly read foreign language input sentences efficiently This can be easily performed.

In this case, the translation level given to the word dictionary is changed from the importance mark given to the normal dictionary.
First information about the degree of importance of the importance mark, the number of words assigned to the dictionary, the second information about the number of words constituting each word of the input sentence, and the second information about the type of part of speech. 3 information and 4th type of word dictionary such as basic dictionary, user dictionary and specialized dictionary
The information is set based on at least one of the information and the like.
A more detailed and accurate translation is made, so that necessary words are correctly translated and not unnecessary words.

In setting the translation level based on the number of word characters, not only is the setting based on whether the number of word characters exceeds a certain number of characters, but also the configuration of each word such as the prefix and suffix of each word. If the setting is made based on whether or not the meaning of each word can be guessed, a more detailed and clear translation is made.

In setting the translation level based on the type of part of speech, the translation level of a specific part of speech is set low, and proper nouns and auxiliary verbs are excluded from the translation target. As a result, it is possible to more efficiently and easily read a foreign language input sentence quickly or roughly.

Further, in the translated word output determination processing, it is determined whether or not to provide a translated word based on linguistic information obtained during at least one of the syntax analysis processing, the syntax conversion processing, and the translated sentence generation processing. And more precise translations are possible.

The linguistic information is the syntax structure of the sentence obtained during the syntax analysis process, the character type configuration of the translated word obtained during the syntax conversion process, and whether the translated word obtained during the syntax conversion process is specified in the field. No, and at least one of the distances between the positions where the same translation is given, etc.
The more information you have, the more accurate the words you need to translate,
In addition, a more detailed and accurate translation that does not translate into unnecessary words is performed.

It is also possible to determine whether or not to add a translation based on the distance between the same translations added immediately before, and to determine whether the number of times of the same translation obtained during the syntax conversion processing is within a predetermined range of the document. If the standard is exceeded, it is necessary not to assign the same translation after that, and it is necessary to assign translations that have been assigned once and remain in human memory or translations that can be easily referenced within the field of view As a result, the translation state is clear, and it is possible to more efficiently and easily read a foreign language input sentence quickly and roughly.

[0043]

Embodiments of the present invention will be described below.

FIG. 1 is a block diagram showing a basic configuration of a translation machine translation apparatus according to an embodiment of the present invention.

In FIG. 1, an input means 1 for inputting a character string and a symbol is connected to an output means 3 via a translation CPU 2 and outputs a translated word corresponding to a word necessary for translation to the output means 3 together with an input sentence. Then, it is displayed or printed by the output means 3 and output. The translation CPU 2 is connected to the storage unit 4, and stores dictionary data and processing results required for translation processing by the translation CPU 2 in the storage unit 4.

The configuration of the translation CPU 2 will be described. The translation level setting unit 5 to which the input unit 1 is connected specifies the translation level according to the language level of the user via the input unit 1 by the user. Further, the field designating means 6 to which the input means 1 is connected allows the user to designate a translation specialty field via the input means 1. The dictionary lookup / morpheme analysis means 7 to which the translation level setting means 5 and the field designation means 6 are connected draws a word dictionary, divides an input sentence into morpheme strings (word strings), and Obtain grammatical information such as part of speech and its translations, and analyze the tense, personal name, number, etc. of each word. This dictionary lookup / morphological analysis means 7
Is connected to the grammatical information, and determines the sentence structure such as the dependency between words in each morpheme sequence of the input sentence using a so-called structure analysis tree. The syntax conversion means 9 to which the syntax analysis means 8 is connected converts the structure of the syntax analysis tree for the input sentence into a syntax analysis tree for the translated sentence, further checks the semantic consistency, and converts the translation for the word. select. The translated sentence generating means 10 to which the syntax converting means 9 is connected determines the arrangement order of the translated words obtained by the syntax converting means 9 and adjusts the utilization change to generate a translated sentence. The source language translation correspondence storage unit 11 to which the translation sentence generation unit 10 is connected performs morphological analysis, syntax analysis, syntax conversion, and translation processing for generating a translated sentence, and then performs translation processing for each word of the input sentence and the translated word. The correspondence is stored and the positional relationship is stored.
The translation output determining means 12 to which the translated sentence generating means 10 is connected determines whether or not to provide a translated word based on the relationship between the translation level specified by the translation level setting means 5 and the translation level of the original word in the word dictionary. Is determined. Translation word output position count storage means 13 to which translation output determination means 12 is connected
Stores the number of output positions of the translated word. Further, an output format shaping unit (corresponding to an original language translation word) to which the translation output determining unit 12 is connected.
The output unit 14 is connected to the output unit 3, and outputs the original sentence and the translated sentence to the output unit 3 in an easily recognizable format. In the case of the present embodiment, the output format shaping means 14 also functions as an original language translated word corresponding output means. After performing the translation processing up to the translation sentence generating means 10, the translation output determining means 12 The translation is output together with the input sentence by associating each word of the sentence with its translation.

The structure of the storage means 4 will be described. The storage means 4 stores grammatical information such as the part of speech of a word and a translated word, and a word dictionary 4a in which level information is provided for each word heading. A translation level buffer 4b for storing the translation level designated by the user, an original language buffer 4c for associating the word number of the original sentence with the original language, and a translated word buffer 4d for associating the word number of the translated sentence with the translated word. An original language translation word correspondence information buffer 4e for storing information as to the number of the translation word corresponding to each source word in the input sentence from the beginning of the normal translation sentence;
4 and an output format buffer 4f for storing the output format of the original sentence and the translated sentence.

Next, a specific hardware configuration of the translation machine translation device will be described with reference to FIG. This translation machine translation apparatus includes a main CPU 15 for controlling each unit, a main memory 4A and a dictionary memory 4a as components of the storage unit 4, and a CRT (cathode ray tube) or LCD (liquid crystal display) as an output unit 3. Display means 3A
, A keyboard 1A as the input means 1, a translation module 16 for performing translation, and a bus line 17 for connecting these components.

As shown in FIG. 3, the translation module 16 translates a source language input via the bus line 17 according to a predetermined translation program and outputs the translated language to the bus line 17 as a target language.
A translation program memory 16A for storing a translation program executed by the translation CPU 2, a buffer A for storing the textual input source language for each word by referring to the dictionary data stored in the dictionary memory 4a A buffer B for storing information such as the part of speech and translation of each obtained word, a buffer C for storing information about the source language structure analysis tree, and a buffer for storing a target language structure analysis tree converted from the structure analysis tree D and a buffer E for storing a sentence prepared in the form of the target language by supplementing an appropriate auxiliary word to the structure analysis tree of the target language, for example, particles or auxiliary verbs in Japanese.

Here, machine translation will be described.

As shown in FIG. 4, there are various analysis levels. When the source language is input via the keyboard 1A at the upper left of FIG.
The morphological analysis of 2, the syntax analysis of level L3,... Machine translation is roughly divided into the following two types depending on the level of analysis processing to be performed. One of them analyzes the intermediate language, which is a concept independent of both the source language and the target language of level L6, and generates a context of level L7, a semantic generation of level L8, a syntactic generation of level L9, and a level L9. This is a pivot method in which morpheme generation and generation of L10 are performed to generate a target language. The other is to perform an analysis up to one of the morphological analysis at the level L2, the syntax analysis at the level L3, the semantic analysis at the level L4, and the context analysis at the level L5 to obtain the internal structure of the source language. This is a transfer method in which a target language is generated after conversion into an internal structure of a target language at the same level as the obtained internal structure of the source language.

Hereinafter, the translation operation of the present invention will be described with reference to FIGS. Here, an example will be described in which the following English sentence is translated into Japanese and a more accurate translated word is output.

First, the original text read by the input means 1 such as the keyboard 1A is stored in the buffer A as shown in FIG. In this buffer A, the original sentence is separated for each word (each space in English) for dictionary lookup, and each word is separated and stored one character at a time for morphological processing such as ending processing. In FIG. 5, one line corresponds to one word. Under the control of the translation CPU 2 based on the translation program stored in the translation program memory 16A, in the processing of the dictionary lookup / morphological analysis means 5, the translation of each word is performed using the dictionary of the dictionary memory 4a in accordance with the original sentence stored in the buffer A. And grammatical information such as the part of speech for the word is obtained and stored in the buffer B. For example, part-of-speech information, which is a part of each information, is stored as shown in FIG. Here, “will”, “her”, and “to” are many parts of speech, but are uniquely determined in the processing of the parsing unit 8 of the translation program. In this way, in dictionary lookup and morphological analysis by the dictionary lookup / morphological analysis means 7,
The dictionary in the dictionary memory 4a is consulted, the input sentence is divided into morpheme strings (word strings), grammatical information such as part of speech and translations for each word are obtained, and tense / person / number are analyzed. You.

In the analysis processing by the syntax analysis means 8, a structure analysis tree indicating a dependency relationship between each word in the word string according to the dictionary data stored in the dictionary memory 4a and the grammatical rules, for example, as shown in FIG. And stored in the buffer C.

The structure analysis tree is determined as follows. In other words, from the grammar rules of the dictionary memory 4a, the sentence → main part, predicate main part → noun phrase predicate → auxiliary verb, verb, noun phrase, preposition phrase noun phrase → pronoun noun phrase → noun preposition phrase → preposition, determiner, noun Is obtained. This rule indicates, for example, that the first rule is “a sentence is made up of a main part and a predicate”. Hereinafter, a structural analysis tree is determined according to this rule. As described above, in the syntax analysis by the syntax analysis unit 8, the structure of the sentence (structure analysis tree) such as the dependency between the words in each morpheme string is determined.

In the translation program, in the semantic analysis processing in the syntax conversion means 9, the structure of the syntax analysis tree for the input English sentence is determined as shown in FIG. 7 using the tree structure conversion rules of the dictionary memory 4a. Then, as shown in FIG. 8, it is converted into a parse tree for Japanese sentences. The result obtained in this way is stored in the buffer D. At the time of the conversion of the parse tree, by using the semantic information in the word dictionary of the dictionary memory 4a, semantic checking is performed between the main words of the parse tree in FIG. Select

In the word dictionary of this embodiment, the following (Table 1)
Is stored.

[0058]

[Table 1]

As shown in Table 1, the part of speech of "take" is a "verb", and its translation has several meanings such as "take", "receive", and "capture". Translation term selection conditions are stored in the semantic information for the verb. Specifically, it indicates that "if the object of the verb is a noun of the semantic category listed there, the translation is taken".

As shown in (Table 1), "child"
Is a “noun”, and the meaning information of the noun stores a semantic category for the translated word. It can be seen that the meaning category of the translated word "child" is "person".

In this example sentence, “tak
It turns out that the object of "e" is "child".
Since the semantic category of the translation of "child" of "d" is "person", when looking for a translation that includes "person" in the translation selection condition of "take", "take along" is found. , "Take" and "child" are translated to "take" and "child", respectively.

As described above, in the semantic analysis by the syntactic conversion means 9, a semantically correct one and a non-semantically correct one are discriminated from a plurality of syntactic analysis results, and a translated word for each word is selected.

In the context analysis processing in the translated sentence generation means 10 of the translation program, the Japanese character “I, her, children, zoo, and the zoo” obtained by the syntax conversion means 9 are taken. The appropriate particle "wa", "no", "ni", "wo" or an auxiliary verb is added to "", and the result is stored in the buffer E in a Japanese form as shown in FIG. This Japanese sentence "I take her child to the zoo. Are output from the translation module 16 and stored in the main memory 4A, and can be displayed on the display device 3A as the output means 3.

As described above, in the context analysis by the translated sentence generation means 10, the topic is understood, and the omission and ambiguity are removed.

At this time, the original language translation word correspondence storage means 11 is started under the control of the translation sentence generation means 10, and the number of the translation word corresponding to each of the original words in the input sentence from the beginning of the normal translation sentence is determined. Is set in the source language translation information buffer 4e as the translation word number. Thereafter, the output format shaping means 14
Then, based on the data in the original language buffer 4c, the translated word buffer 4d, and the original language translated word correspondence information buffer 4e, the original language (word)
Is stored in the output format buffer 4f, and is output to the display device 3A or the like as the output means 3 in a translation format in which the required words are provided with the translated words.

Next, the operation of the present embodiment will be described for the case where a translation is not given to a word that does not require a translation according to the language level of the user, and a translation is given to the minimum required word desired by the user. This will be described in detail with reference to the block diagrams of FIG. 1 (Tables 2 to 4).

First, it is assumed that the following sentence has been input to the input means 1.

The earth contains a large number of m
This input sentence is subjected to a dictionary creation / morphological analysis, syntax analysis, and syntax conversion, and a translated sentence generation process is performed in the same manner as in the case of the translation machine translation of the present invention described above.

At this time, the original language translation word correspondence storage means 11 is started under the control of the translation sentence generation means 10, and the number of the translation word corresponding to each of the original words in the input sentence from the beginning of the normal translation sentence is determined. Is set in the source language translation information buffer 4e as the translation word number. The following (Table 2) shows an example of the source language translation information buffer 4e. Incidentally, the stored contents of the translated word (translated word buffer 4d) and the original word (original word buffer 4c) corresponding to the respective numbers are shown in the following (Table 3) and (Table 4), respectively.

[0070]

[Table 2]

[0071]

[Table 3]

[0072]

[Table 4]

Thereafter, the output format shaping means 14 converts the character string corresponding to the original word (word) to the translated word on the basis of the data of the original word buffer 4c, the translated word buffer 4d and the translated word correspondence information buffer 4e. Stored in
The translated word is output together with the input sentence as an output format shaping result by the display device 3A or the like as the output means 3 as shown below.

The translated word output judging process in the case where the translation is performed in this manner is described in accordance with the flowchart of FIG. (Table 5).

In the dictionary memory 4a of the word dictionary, each heading is assigned a level based on level information such as difficulty level, frequency level, and importance level. In this embodiment, four levels of “advanced / intermediate / beginner / extreme” are assumed. The following (Table 5)
2 schematically shows the basic dictionary contents leveled for each heading unit of each word in the input sentence example.

[0076]

[Table 5]

In this (Table 5), for example, “are”
Is a basic word, so "beginner" is given,
No translation is needed for the word "super beginner". The user uses the translation level setting means 5 to set the translation level as his / her own academic ability level. In this embodiment, three levels of “advanced / intermediate / beginner” are assumed. The designated level information is stored in the translation level buffer 4b.

First, the case where the translation level is “intermediate” will be described. The same input sentence as the previous input sentence is input, and a word-by-word translation for each source word is obtained in the same manner, and stored in the source word buffer 4c, the target word buffer 4d, and the source word translation word correspondence information buffer 4e. It is assumed that

As shown in FIG. 10, the translation output judging means 12 first focuses on the first word "the" in step s1. In step s2, word level information is extracted. As shown in Table 5 schematically showing the basic dictionaries leveled for each heading unit, it can be seen that the dictionary is "super elementary".

Since the translation level is "intermediate" and does not constitute an idiom, NO is obtained in steps s3 and s5, and YES is obtained in step s9. In step s10, it is determined whether the level information of the word "the" is "super elementary" or "elementary". The level information of the word “the” is “super elementary”, and the translated word of “the” which is a word below the “elementary” is not output in step s11. Further, the process proceeds to step s16, where it is determined whether or not the processing of all the words of the input sentence has been completed. If all words have not been processed, step s17
Attention is paid to the next word "earth". When the level information of the word is extracted in step s2, the word is "beginner". Therefore, similarly to the case of the head word "the", no translated word is output in step s11.

Similarly, attention is paid to the next word “contain” in step s17, and when the level information is extracted in step s2, it is “intermediate”. In step s10 through step s9, this is “intermediate”. It is determined, and the translated word of "contain" is output in step s12.

This actual output is output to the output format shaping means 14.
The output format of the original and translated words is stored in the output format buffer 4f, and output by the output means 3 in a format such as display or printing. The output state of the output means 3 at this time is as follows.

[0083] Further, in step s17, attention is paid to the next word "a", level information is extracted in step s2, and in step s3, it is determined whether or not the phrase after "a" is an idiom of "a large number of". If there is, the translated word is output irrespective of the translation level specified by the user and the level information of the word. The output state of the output means 3 at this time is as follows.

[0084] Thereafter, the process proceeds in the same manner, and eventually, at the intermediate level, the following output result is obtained.

[0086] Next, regarding the processing when the user designates the "advanced" translation level, the above-mentioned "intermediate" The differences from the above case will be described.

Although the words “the” and “earth” are not output in the same manner as described above, the next word “contain” is “intermediate”.
On the other hand, since the translation level specified by the user is "advanced", it is determined in step S14 that "do not output". Similarly, the word "metal" of the level information "intermediate"
And "useful" are not output, and eventually, at the advanced level, only the idiom of "a large number of" is output, and the output state of the output means 3 is as follows.

The output format shaping results when three translation levels specified by the user are set as described above are shown below.

(Early Level Example) The earth contains a large number of metals which are useful to man. The earth including many metals is useful (an intermediate level example) The earth contains a large number
of metals which are used to man. Including many metals Informative (advanced level example) The earth contains a large number
of metals which are used to man. A large number of specific examples 1 to 4 of the criterion for assigning the level information of each word stored in the word dictionary will now be described. Using the reference of the level information, the translated word output determining unit 12 determines whether or not to output the translated word.

(Specific Example 1) Dictionary Memory 4a in Word Dictionary
Is set based on how many of the English-Japanese dictionaries have the importance mark added thereto. In other words, a word with many dictionaries to which a high importance mark is given is a basic word, and a low word level is given because many people will know the translation and meaning.

As described above, of the importance marks assigned to the normal dictionary, based on the information on how many importance marks are assigned to the importance marks and how many dictionaries are assigned to the dictionaries. , Dictionary memory 4 in word dictionary
The level information of the translation given to a is set.

(Specific Example 2) Dictionary Memory 4a in Word Dictionary
The level information of each word to be assigned to is determined by the character configuration of the heading. For example, a word whose heading is composed of 11 or more characters is given a high word level because it is a word that is difficult to understand and memorize. For example, the following two words have a character configuration of 11 or more characters, and a translation is added to these two words.

Supernatural Supercalender Supernatural paper is made glossy. Even if the word has a character composition of 11 characters or more, it is a word whose prefix or suffix is concatenated and lengthened in terms of word composition. In addition, the word level is set low for those which can easily be deduced from the meaning of the fragments or the meaning of the whole word. For example, both of the above two words have a prefix of “super”, and the other part is a basic word for both “natural” and “calender”. While the latter "calender" has a completely different meaning. Therefore, the former word level is set low so that no translation is given, and the latter word level is set high so that the translation is given.

Note that most idioms have more than 11 characters, so if this criterion is used, the word level (complex word level) of the idiom will be set to a higher word level. . In this case, for example, the word level (compound word level) of the easy idiom may be set by giving a low word level.

As described above, the translation level information to be assigned to the dictionary memory 4a in the word dictionary is set based on the information on the number of words constituting each word of the input sentence. Level information is set based on whether or not the number of words exceeds a certain number of characters, and whether or not the meaning of each word can be inferred by the configuration of each word such as the prefix and suffix of each word.

(Specific Example 3) Dictionary Memory 4a in Word Dictionary
The level information of each word to be assigned to is determined according to the type of part of speech. That is, a specific word class is uniformly assigned a low word level. For example, the auxiliary verb is a part of speech that is not subject to translation because of its small number. Similarly, since proper nouns often become katakana translations or erroneous translations, no translation is given.

As described above, the translation level information to be assigned to the dictionary memory 4a in the word dictionary is set based on the information on the type of part of speech. That is, the specific part of speech is set at a low translation level, and proper nouns and auxiliary verbs are excluded from the translation target.

(Specific Example 4) Dictionary Memory 4a in Word Dictionary
The level information of each word to be assigned to is determined according to which type of dictionary has been registered. That is, the word registered in the basic dictionary is likely to be a basic word, and the word level is set low. On the other hand, a word registered in a user dictionary, a specialized dictionary, or the like is set to a high word level because the word itself or its translation may be special.

As described above, the translation level information to be assigned to the dictionary memory 4a in the word dictionary is set based on the information on the type of the word dictionary such as the basic dictionary, the user dictionary, and the specialized dictionary.

Next, instead of the predetermined level information given to the dictionary memory 4a in the word dictionary as described above, the above-mentioned translation (translation) processing (syntax analysis processing, syntax conversion processing and translated sentence generation processing) Specific examples 5 to 7 of the assignment criteria for the translated word output determining means 12 to determine whether to assign a translated word based on the linguistic information obtained therein will be described.

(Specific Example 5) By using the syntax structure of the original sentence obtained as a result of the syntax analysis, the translated word output determining means 12 makes an output determination for determining whether or not to add a translated word. That is, after the syntactic analysis, as shown in FIG. 7, it is possible to obtain a structure serving as a skeleton of a sentence such as a noun phrase corresponding to the main part or a main verb of a predicate. By using this parsing information, even if the word level is low, a word that plays an important role in the sentence is given a translation. In particular, at the elementary level, words that play the role of subject or predicate are assigned a translation at any level, and conversely, at the advanced level, modifiers (including words in relative clauses, etc.) No translation is given to the word in parentheses).

Explaining specifically the following input sentence example, at the "elementary" level, not only the judgment at the word level but also the structure of sentences such as S (subject), V (verb), O (object), etc. Translations are also given to the elements. At the “intermediate” level, which is determined only at the word level, no translated word is given only by the constituent elements of the sentence. For example,
The translation of “friend”, which is a component (subject) of the sentence, was given at the “beginner” level, but was not given at the “intermediate” level. Conversely, at the “advanced” level, if the determination is made only at the word level, translations such as “derived” that should be given are also suppressed because they are in the relative clauses as modifiers.

(Beginner level: Judgment at the word level and sentence components.) Her friends extolled Japan where technology is derived from the theory. Friends acclaimed technology Derived theory (Intermediate level: Judgment at the word level.) Her friends extolled Japan where technology is derived from the theory ・ acclaimed technology derived theory (advanced level: no translation is given to modifiers such as relative clauses) Her friends extolled Japan where technology is derived from the theory ・ acclaimed In addition, the translation output determination unit 12 determines whether or not to add a translation based on not only the word level but also the sentence syntax structure (sentence constituent element, modifier, etc.) obtained during the syntax analysis processing.

(Specific Example 6) The most appropriate translated word is determined by selecting a translated word using a semantic process performed during syntax conversion.
The output is determined by determining the type of the translated word. For example, a translated word in which the pronunciation of an English word is converted into katakana as it is is not useful information for deciphering an English sentence, and therefore, no translated word is added at any level.

An example of an input sentence shown below will be specifically described.

(Before Katakana translation is suppressed) The text describes an mechanism and programmig model for new software. The text describes an mechanism and programmig model for new software. In the translation results after Katakana translation suppression,
It turns out that it is easy to see clearly.

As described above, the translation output determining means 12 determines whether or not to add a translation based on the character type configuration of the translation obtained during the syntax conversion processing, not only at the word level.

(Specific Example 7) Even for the same word, a translated word may not be necessary depending on the field. As a practical machine translation system,
The user specifies the field of the document to be translated by the field specifying means 6 via the input means 1. This is for performing translation proper selection, translation output determining unit this field information 12
It is also used for the determination of translation word assignment performed in. That is, when a translated word (special translated word) specified in the field is selected, the translation output determining means 12 determines that the translated word is added even if the level information of the word is low.

For example, in the first output judgment result of the following example sentence, no translation is given because “attack” has a low word level, but the “medical” field is specified in the second output judgment result. In this case, the selected word "illness" is translated.

In this way, the translation output determining means 12 determines whether or not the translated word obtained during the syntax conversion processing is a field-specified word. It is determined whether or not to add a translation based on the information.

(Specific Example 8) In this specific example 8, a description will be given of an assignment criterion in which the translation output determination means 12 determines whether or not to add a translation word based on information from the translation output position count storage means 13.

Each time a translated word is added, the translated word and its display position are stored in the translated word output position count storage means 13. Based on the information from the translated word output position number storage unit 13, the translated word can be added only when the translated word is at least a certain distance from the previously displayed position.

If the number of times a given translation is given exceeds a certain standard within a certain range of the document, the translation is not added thereafter. It is also possible to determine whether or not to assign a translation to all words (translations) based on the distance from the position immediately before the translation.

In the first translation result of the following example sentence (before this rule is added), the critical temperature (critical temperature), ca
Words that appear multiple times, such as rbon (carbon), heat (heat), and temperature (temperature), are also translated each time, but the second translation result of the example sentence (after adding this rule) The figure shows a case where the translated words appearing after the second time are suppressed, and it can be seen that the translation is easy to see clearly.

(Before this rule is added) In the first place, steel which contains very little carbon will be firstly contained carbon milder than steel. Secondly, we can heat the carbon above a certain mildly secondarily heated carbon critical temperature.At this critical temperature, changes begin critical temperature to take place in the molecular structure.In the process known as molecular structure annealing, we heat the carbon above the critical temperature.We heat it Critical temperature heating up beyond the critical temperature, and then quench it in other liquid.The temperature drop fixes the change in the carbon which occurred temperature carbon occurs at the critical temperature.But a bar of this hardened carbon is more critical temperature hardened carbon liable to fracture than normal carbon.We therefore heat it again to a prone to fracture normal carbon heating temperature bel In the first place, steel which contains very little carbon will be carbon in the first place, mildly than steel.Secondly, we can heat the carbon above a certain mild ow the critical temperature. Critical temperature.At this critical temperature, changes begin Critical temperature to take place in the molecular structure.In the process known as Molecular structure annealing, we heat the carbon above the critical temperature.We heat it beyond the critical temperature, and then quench it in other liquid.The temperature drop fixes the change in the carbon which occurred temperature occurs at the critical temperature.But a bar of this hardened carbon is more liable to fracture than normal carbon. We therefore heat it again to a tendency of crushing Normal temperature below the critical temperature. The operation of the power determining means 12 will be described with reference to the flowchart of FIG.

As shown in FIG. 11, first, in step S1
Focus on the first word. Next, in step S2, the level information, the part of speech, and the translated word of the focused word are extracted. Further, it is determined whether the part of speech of the focused word is an auxiliary verb or a proper noun in step A1, and if YES, that is, if the focused part of speech of the word is an auxiliary verb or a proper noun, no translated word is output in step A2. In step S16, it is determined whether or not all the words have been noticed. If not all the words have been noticed, the next word is noticed in step S17, and the process proceeds to step A1 via step S2.

Further, if the part of speech of the word focused in step A1 is not an auxiliary verb or proper noun, step A3
Then, it is determined whether the word of interest is a user dictionary or a specialized dictionary, or whether the translated word is a translated word in a specified field. If YES, the translated word is output in step A4. Further, it is determined in step S16 whether or not all the words have been noticed.
At 17, the next word is noted, and the process proceeds to step A1 via step S2.

At step A1 and step A3, N
In the case of O, the process proceeds to step A5, and in step A5, it is determined whether or not it is sufficiently distant from the immediately preceding translated word. If it is not sufficiently distant from the immediately preceding word with the same translation, the translation is not output in step A6. further,
In step S16, it is determined whether or not all the words have been noticed. If not all the words have been noticed, the next word is noticed in step S17, and the process proceeds to step A1 via step S2.

Further, if NO in steps A1 and A3 and YES in step A5, step S
Then, in step S5, it is determined whether the translation level is "elementary". If the translation level is "elementary", it is determined in step S6 whether the word of interest has the level information "super elementary" and is a modified part of the sentence. If the level information of the word is “super elementary” and the word is a modifier of the sentence, the translated word is not output in step S7. If the level information of the word is “beginner” and the word is not a modification part of the sentence, the translated word is output in step S8. Further, it is determined whether or not all the words have been focused on in step S16. If not all the words have been focused on, the next word is focused on in step S17, and the process proceeds to step A1 via step S2.

From step A1 to step S7
When the process proceeds to step S7, the translation of the word of interest in step S7 is
It is determined whether or not it is composed only of katakana. If the translated word is composed only of katakana, the translated word is not output in step A8. Further, it is determined in step S16 whether or not all the words have been noticed.
At 17, the next word is noted, and the process proceeds to step A1 via step S2.

Further, the process proceeds from Step A1 to Step S7 in the same manner. If the translated word is not Katakana but NO and the process proceeds to Step S9, it is determined in Step S9 whether the translation level is "intermediate". If the translation level is "intermediate", it is determined in step S10 whether the level information of the word of interest is "super elementary" or "elementary". If the level information of the word is "beginner" or "beginner", the translated word is not output in step S11. If the level information of the word is not "beginner" or "beginner", the translated word is output in step S12. Further, step S
At 16, it is determined whether or not all the words have been noticed. If not all the words have been noticed, the next word is noticed at step S 17, and the process proceeds to step A 1 via step S 2.

Further, the process proceeds from step A1 to step S9 in the same manner. If the translation level is not “intermediate” in step S9 and the process proceeds to step S13 with NO, the word of interest in step S13 is the level of the word. It is determined whether the information is “advanced” and the word is the SVO part of the sentence. If the level information of the focused word is “advanced” and the word is the SVO part of the sentence, the translated word is output in step S15. Further, step S16
It is determined whether or not all the words have been noticed. If not all the words have been noticed, the next word is noticed in step S17, and the process proceeds to step A1 via step S2.

Further, similarly from step A1, the process proceeds to step S13. If the level information of the word is "advanced" in step S13 and the word is not the SVO part of the sentence but NO, the translated word is output in step S14. Instead, it is determined in step S16 whether all the words have been noticed. If all the words have been noticed, the translation process ends.

As described above, the determination as to whether or not to add a translated word is made finely, and the translated word is more accurately suppressed for a word that does not require a translated word according to the level of the user. Can read fast or roughly.

As described above, in the present invention, the sentence input from the input means 1 is divided into a morpheme sequence of words, and a word dictionary is used for each word, and the tense, personal name, number, etc. of the word are determined. To analyze. The syntax analysis unit 8 determines the input sentence structure such as the dependency between words in this morpheme sequence,
The syntax conversion means 9 converts this into a sentence structure for the translated sentence, checks semantic consistency, and selects a translated word for the word. Further, the translated sentence generating means 10 determines the arrangement of the translated words, adjusts the utilization change, and generates a translated sentence. As described above, after performing the morphological analysis, the syntax analysis, the syntax conversion, and the translation processing for generating the translated sentence, the output format shaping unit 14 performs translation by associating each word of the input sentence with the translated word.
In this translation, the translated word output determining unit 12 determines whether or not to provide a translated word, based on the relationship between the translated level set by the translated level setting unit 5 and the translated level in the word dictionary. This makes it possible to efficiently and easily read a foreign language input sentence quickly or roughly.

[0125]

As described above, according to the present invention, after performing a morphological analysis, a syntax analysis, a syntax conversion, and a translation process for generating a translated sentence, each word of the input sentence is corresponded to the translated word and translated. , It is possible to output a translation with higher accuracy, and the user can correctly understand the input sentence,
Fast and rough reading of foreign language input sentences can be performed efficiently and easily.

[0126] Level information is provided for each word heading. The user designates a translation level, and a translation is given according to the relationship between the designated translation level and the level information of the original word (word). Since it is determined whether or not the word is unnecessary for the translation according to the level of the user, the user can efficiently read the input sentence quickly or roughly.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a basic configuration of a translation machine translation apparatus according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a specific hardware configuration of the translation machine translation device of FIG. 1;

FIG. 3 is a block diagram showing a detailed configuration of a translation module 16 in FIG. 2;

FIG. 4 is a diagram schematically illustrating the concept of machine translation.

FIG. 5 is a diagram schematically showing stored contents of a buffer A of FIG. 3;

FIG. 6 is a diagram schematically showing the contents stored in a buffer B of FIG. 3;

FIG. 7 is a diagram schematically showing stored contents of a buffer C of FIG. 3;

FIG. 8 is a diagram schematically showing the contents stored in a buffer D of FIG. 3;

FIG. 9 is a diagram schematically showing the contents stored in a buffer E of FIG. 3;

FIG. 10 is a flowchart showing an example of the operation of the translated word output determining means 12 of FIG. 1;

FIG. 11 is a flowchart showing another example of the operation of the translated word output determination means 12 of FIG. 1;

FIG. 12 is a block diagram showing a configuration of a conventional translation support device.

FIG. 13 is a block diagram illustrating a configuration of a conventional machine translation device.

[Explanation of symbols]

 Reference Signs List 1 input means 1A keyboard 2 translation CPU 3 output means 3A display means (display device) 4 storage means 4A main memory 4a dictionary memory 4b translation level buffer 4c original word buffer 4d translation word buffer 4e original language translation word correspondence information buffer 4f output format buffer 5 translation Assignment level setting means 6 Field designation means 7 Dictionary lookup / morphological analysis means 8 Syntax analysis means 9 Syntax conversion means 10 Translation sentence generation means 11 Original language translation word correspondence storage means 12 Translation word output determination means (translation output determination means) 13 Number of translation word output positions Storage means 14 output format shaping means 15 main CPU 16 translation module 16A translation program memory AE buffer

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G06F 17/27-17/28

Claims (4)

(57) [Claims] (1) grammatical information such as a part of speech for a word;
The translations are stored, and a translation level is assigned to each translation.
A given word dictionary and an input sentence are divided into word strings.
When you get grammatical information such as part of speech for each word and translated words
Both are dictionary lookups for analyzing the tense, person, and number of each word.
The morphological analysis means and the grammatical information obtained by the dictionary lookup / morphological analysis means
The input sentence structure such as the dependency relationship between the words in the word string.
And parsing means for determining the granulation, versus an input sentence structure determined by said syntax analysis means translation
Into a sentence structure, and check for semantic consistency.
Syntactic conversion means for selecting the translation for the word; and determining the order of the translations obtained by the syntax conversion means.
A translation sentence generating means for adjusting the usage change to generate a translated sentence.
A translation machine translation device provided with a user, wherein the user can determine the degree of difficulty, frequency or importance of each word, etc.
Specify the translation level set based on the level information of
Possible translation level setting means, and a translation level designated by the translation level setting means.
Translated words according to the translation level in the word dictionary
After performing translation processing up to the translation output determining means and the translation sentence generating means for determining whether or not to add
The translation of the input sentence
Correspondence between each word and its translation is used to define the translation as the input sentence.
Output source for original language translations that output both in an easy-to-view format
And the translation output determination means, wherein the translation level setting means
Specified translation level and translation in the word dictionary
Determine whether to add a translation word based on the relationship with the level
Along with the syntax structure of the sentence obtained during the parsing process,
The character type configuration of the translated word obtained during the conversion process;
Whether the translation obtained during the process is specified in the field
Or the distance between the positions where the
At least one of the pieces of information is linguistic information, and based on the linguistic information,
Determining whether or not to add a translated word based on the
That translation pretend machine translation apparatus. 2. The distance between the positions where the same translation is given is
Characterized in that the distance between the same translation word imparted before
The translation machine translation device according to claim 1 . 3. grammatical information such as part of speech for a word and
The translations are stored, and a translation level is assigned to each translation.
A given word dictionary and an input sentence are divided into word strings.
When you get grammatical information such as part of speech for each word and translated words
Both are dictionary lookups for analyzing the tense, person, and number of each word.
The morphological analysis means and the grammatical information obtained by the dictionary lookup / morphological analysis means
The input sentence structure such as the dependency relationship between the words in the word string.
And parsing means for determining the granulation, versus an input sentence structure determined by said syntax analysis means translation
Into a sentence structure, and check for semantic consistency.
Syntactic conversion means for selecting the translation for the word; and determining the order of the translations obtained by the syntax conversion means.
A translation sentence generating means for adjusting the usage change to generate a translated sentence.
A translation machine translation device provided with a user, wherein the user can determine the degree of difficulty, frequency or importance of each word, etc.
Specify the translation level set based on the level information of
Possible translation level setting means, and a translation level designated by the translation level setting means.
Translated words according to the translation level in the word dictionary
After performing translation processing up to the translation output determining means and the translation sentence generating means for determining whether or not to add
The translation of the input sentence
Correspondence between each word and its translation is used to define the translation as the input sentence.
Output source for original language translations that output both in an easy-to-view format
And the translation output determination means, wherein the translation level setting means
Specified translation level and translation in the word dictionary
Determine whether to add a translation word based on the relationship with the level
Along with the syntax of the sentence obtained during the parsing process.
Structure, the character type configuration of the translated word obtained during the conversion process,
The translation obtained during the conversion process is specified in the field
Or not, and the distance between the positions where the same translation is given,
At least one of the information is linguistic information, and the linguistic information
To determine whether or not to add a translation based on the
The number of times the same translation is obtained during the syntax conversion
If the specified standard is exceeded within the specified range,
Pretend machine translation apparatus mean that the feature that it does not grant one translation. 4. A translation level assigned to the word dictionary.
Is the first information about the importance mark given to the normal dictionary.
And a second number relating to the number of word characters constituting each word of the input sentence.
Information, third information on the type of part of speech, and dictionary types such as a basic dictionary, user dictionary, or specialized dictionary
Fourth information, is composed of at least one of information of, when setting the translation swing level based on the first information regarding,
Among several dictionaries, several dictionaries are assigned importance marks.
And how important the importance mark is
Set based on whether the chromatography click, when setting the translation swing level based on the second information,
Whether the number of word characters exceeds a certain number of characters, or
The structure of each word, such as the prefix and suffix of each word
When setting the translation level based on whether or not the meaning of each word can be estimated, and setting the translation level based on the third information,
Set the level lower for specific parts of speech,
When the auxiliary verb is excluded from the translation target and the translation level is set based on the fourth information,
Words registered in the basic dictionary lower the word level,
Words registered in the user dictionary or specialized dictionary
By setting the word level higher, a translation level to be given in the word dictionary is set.
4. The translation machine translation device according to claim 1, wherein: